1 workshop Barcelona, April 22, 2015
2 The mPlane community Many Academic and research institutions come from the Traffic Monitoring Analysis (TMA) COST Action: PoliTo, ETH, ENST, Eurecom, UniLiege, FTW, NEC Some others were part of previous EU projects: PRISM: FTW, ETH NapaWine: PoliTo, ENST, NetVisor, NEC DEMONS: FTW, ETHZ, NEC, TID … A community of “measurement” experts
3 The nowadays Internet “The Internet is the first thing that humanity has built that humanity doesn't understand, the largest experiment in anarchy that we have ever had.” Eric Schmidt – ex Google Exec. Chairman
Network neutrality and measurements… Marco Mellia Politecnico di Torino I’m by far not a network neutrality expert… … I may be a measurement expert…
5 Network neutrality Net neutrality (also network neutrality, Internet neutrality, or net equality) principle: Internet service providers and governments should treat all data on the Internet equally No discrimination or charging differentially by user, content, site, platform, application, type of attached equipment, or mode of communication
Example of measurements from the ISP point of view
7 Something we “watch” 5-May6-May7-May8-May9-May10-May SUBNETNAME with AIRPORT code#flowTru avg#flowTru avg#flowTru avg#flowTru avg#flowTru avg#flowTru avg fra02s08.c.youtube.com fra02s15.c.youtube.com mil01s12.c.youtube.com par08s06.c.youtube.com par08s06.c.youtube.com lhr14s08.c.youtube.com fra07s13.c.youtube.com fra07s19.c.youtube.com par03s06.c.youtube.com par03x04.c.youtube.com mia02s11.c.youtube.com ams09x06.c.youtube.com dfw06x02.c.youtube.com fra07x03.c.youtube.com lhr22s16.c.youtube.com zrh04s03.c.youtube.com mil02s01.c.youtube.com bru02t11.c.youtube.com fra07t13.c.youtube.com lhr22s11.c.youtube.com fra07s03.c.youtube.com ISPs
8 Something we observe 5-May6-May7-May8-May9-May10-May SUBNETNAME with AIRPORT code#flowTru avg#flowTru avg#flowTru avg#flowTru avg#flowTru avg#flowTru avg fra02s08.c.youtube.com fra02s15.c.youtube.com mil01s12.c.youtube.com par08s06.c.youtube.com par08s06.c.youtube.com lhr14s08.c.youtube.com fra07s13.c.youtube.com fra07s19.c.youtube.com par03s06.c.youtube.com par03x04.c.youtube.com mia02s11.c.youtube.com ams09x06.c.youtube.com dfw06x02.c.youtube.com fra07x03.c.youtube.com lhr22s16.c.youtube.com zrh04s03.c.youtube.com mil02s01.c.youtube.com bru02t11.c.youtube.com fra07t13.c.youtube.com lhr22s11.c.youtube.com fra07s03.c.youtube.com
9 Something we observe 5-May6-May7-May8-May9-May10-May SUBNETNAME with AIRPORT code#flowTru avg#flowTru avg#flowTru avg#flowTru avg#flowTru avg#flowTru avg fra02s08.c.youtube.com fra02s15.c.youtube.com mil01s12.c.youtube.com par08s06.c.youtube.com par08s06.c.youtube.com lhr14s08.c.youtube.com fra07s13.c.youtube.com fra07s19.c.youtube.com par03s06.c.youtube.com par03x04.c.youtube.com mia02s11.c.youtube.com ams09x06.c.youtube.com dfw06x02.c.youtube.com fra07x03.c.youtube.com lhr22s16.c.youtube.com zrh04s03.c.youtube.com mil02s01.c.youtube.com bru02t11.c.youtube.com fra07t13.c.youtube.com lhr22s11.c.youtube.com fra07s03.c.youtube.com
10 Something we observe 5-May6-May7-May8-May9-May10-May SUBNETNAME with AIRPORT code#flowTru avg#flowTru avg#flowTru avg#flowTru avg#flowTru avg#flowTru avg fra02s08.c.youtube.com fra02s15.c.youtube.com mil01s12.c.youtube.com par08s06.c.youtube.com par08s06.c.youtube.com lhr14s08.c.youtube.com fra07s13.c.youtube.com fra07s19.c.youtube.com par03s06.c.youtube.com par03x04.c.youtube.com mia02s11.c.youtube.com ams09x06.c.youtube.com dfw06x02.c.youtube.com fra07x03.c.youtube.com lhr22s16.c.youtube.com zrh04s03.c.youtube.com mil02s01.c.youtube.com bru02t11.c.youtube.com fra07t13.c.youtube.com lhr22s11.c.youtube.com fra07s03.c.youtube.com
11 Something we observe 5-May6-May7-May8-May9-May10-May SUBNETNAME with AIRPORT code#flowTru avg#flowTru avg#flowTru avg#flowTru avg#flowTru avg#flowTru avg fra02s08.c.youtube.com fra02s15.c.youtube.com mil01s12.c.youtube.com par08s06.c.youtube.com par08s06.c.youtube.com lhr14s08.c.youtube.com fra07s13.c.youtube.com fra07s19.c.youtube.com par03s06.c.youtube.com par03x04.c.youtube.com mia02s11.c.youtube.com ams09x06.c.youtube.com dfw06x02.c.youtube.com fra07x03.c.youtube.com lhr22s16.c.youtube.com zrh04s03.c.youtube.com mil02s01.c.youtube.com bru02t11.c.youtube.com fra07t13.c.youtube.com lhr22s11.c.youtube.com fra07s03.c.youtube.com
12 Something we observe 5-May6-May7-May8-May9-May10-May SUBNETNAME with AIRPORT code#flowTru avg#flowTru avg#flowTru avg#flowTru avg#flowTru avg#flowTru avg fra02s08.c.youtube.com fra02s15.c.youtube.com mil01s12.c.youtube.com par08s06.c.youtube.com par08s06.c.youtube.com lhr14s08.c.youtube.com fra07s13.c.youtube.com fra07s19.c.youtube.com par03s06.c.youtube.com par03x04.c.youtube.com mia02s11.c.youtube.com ams09x06.c.youtube.com dfw06x02.c.youtube.com fra07x03.c.youtube.com lhr22s16.c.youtube.com zrh04s03.c.youtube.com mil02s01.c.youtube.com bru02t11.c.youtube.com fra07t13.c.youtube.com lhr22s11.c.youtube.com fra07s03.c.youtube.com
13 Something we observe 5-May6-May7-May8-May9-May10-May SUBNETNAME with AIRPORT code#flowTru avg#flowTru avg#flowTru avg#flowTru avg#flowTru avg#flowTru avg fra02s08.c.youtube.com fra02s15.c.youtube.com mil01s12.c.youtube.com par08s06.c.youtube.com par08s06.c.youtube.com lhr14s08.c.youtube.com fra07s13.c.youtube.com fra07s19.c.youtube.com par03s06.c.youtube.com par03x04.c.youtube.com mia02s11.c.youtube.com ams09x06.c.youtube.com dfw06x02.c.youtube.com fra07x03.c.youtube.com lhr22s16.c.youtube.com zrh04s03.c.youtube.com mil02s01.c.youtube.com bru02t11.c.youtube.com fra07t13.c.youtube.com lhr22s11.c.youtube.com fra07s03.c.youtube.com
14 Who to blame when YouTube doesn’t work OK? AS 2 AS 1 ISP Network Devices?ISP?Internet?YouTube? G-CDN
15 Diagnosis #1 change in the cache-selection policy? For every time-bin: compute distribution of volume across all /24 Google CDN subnets Distribution compared with custom Kullback-Leibler metric Comparisons results plotted on heatmap Traffic distribution across /24 subnetsTemporal Similarity Plot (TSP) Conclusion A different server selection policy is deployed exactly on the same day when the anomaly starts occurring! YES!
16 The origin of the anomaly is the cache selection done by Google Diagnosis #2 given that change... who is to blame: new servers or path? Diagnosis #3 is there a correlation between increased RTT and throughput? NO Conclusion Diagnosis #4.... YES/NO
17 Network neutrality Net neutrality (also network neutrality, Internet neutrality, or net equality) principle: Internet service providers and governments should treat all data on the Internet equally No discrimination or charging differentially by user, content, site, platform, application, type of attached equipment, or mode of communication In the previous case, the ISP has no role… … it is a victim (as the user)
18 Network neutrality The network has never been neutral Ethernet is not neutral WiFi is not neutral TCP is not neutral BGP is not neutral ADSL is not neutral …… The network has been designed to be NOT NEUTRAL 20+ years of research on QoS!!! How can we measure the “neutrality”?
19 What about applications? So my daughter cannot watch youtube on her iPad?
20 What about applications? Are applications supposed to be neutral?
And speaking about my daughter…
22 The nowadays Internet “The Internet is the first thing that humanity has built that humanity doesn't understand, the largest experiment in anarchy that we have ever had.” Eric Schmidt – ex Google Exec. Chairman
23 The nowadays Internet If you ask my daughter… “What is the Internet?”…
24 Take 100 users, and observe which services they contact while on the Internet… What Services does My Daughter Use? Google Third-Party sites Take 100 users, and observe which services they contact while on the Internet… … most of those are so called “third-party sites” - aka tracking services… Welcome ScorecardResearch, […] a leading global market research effort that studies and reports on Internet trends and behavior. ScorecardResearch conducts research by collecting Internet web browsing data and then uses that data to help show how people use the Internet, what they like about it, and what they don’t. ScorecardResearch collects data through […] web tagging.
25 Tracking services 233 out of the top-500 hostnames contacted by users are tracking service 196 of this 233 are contacted also using HTTPS To protect my daughter’s privacy? Or to protect their business? First tracker contacted within 1 second for 78% of users 71% of services embed at least one tracker Honestly, I’m not worried about network neutrality… I’m very worried about my daughter's privacy…